Read Product Growth with Scottie
Keep Product Growth. Lose the inbox pileup.
Scottie reads the newsletters you choose, keeps the useful differences, and attaches every original link. Product Growth stays in your reading life without becoming another email to manage.
An independent guide. Scottie isn't affiliated with, endorsed by, or sponsored by Product Growth. The publication owns its name, writing, and subscription terms.
From Scottie · July 31, 2026
See Product Growth in a Scottie brief
Scottie read Product Growth, Lenny's Newsletter, and ByteByteGo Newsletter for a reader with the priorities shown below. Start with the rundown, open the full brief, or check every issue behind it.
3sources
14issues read
6stories included
2 of 4 Product Growth issues included
What shaped this brief
Reader priorities
A product manager building AI features who needs practical guidance on agents, evaluation, distribution, and day-to-day product work.
These are illustrative priorities, not a customer’s data.
Sources in this brief
- Product GrowthThe publication this guide is about
- Lenny's NewsletterAdds broader product leadership, growth, and team-management guidance.
- ByteByteGo NewsletterAdds the system architecture behind agentic and large-scale product patterns.
Scottie
Scottie example brief
July 31, 2026 · Executive brief
01 / The rundown
- Ex-Meta PM Daniel McKinnon introduced a 90-minute cold-start framework that uses AI to generate evaluation sets before shipping agent features.
- OpenAI detailed how persistent WebSockets, prompt prefix caching, and parallel safety checks reduce agent loop latency and lower token costs.
- Search architectures at DoorDash, Instacart, and Uber Eats show how to balance offline graph enrichment against real-time fine-tuned model inference.
Read the complete brief 6 stories · 3 action items
02 / The briefing
01 / main
Build cold-start AI agent evals before shipping features
The takeaway: Skip lengthy PRDs and write an evaluation set before launch. Ex-Meta PM Daniel McKinnon shares a 90-minute cold-start framework that uses AI to construct 100 test prompts, finding your product's floor and ceiling before gathering production data.
Concrete details
- Constructing a 100-prompt evaluation set takes roughly 90 minutes using LLM-assisted generation.
- Evals act as specs, replacing traditional product requirement documents when handing off features to engineering.
Why it matters for this reader: As an AI PM, owning evals gives you precise quality targets for engineering teams, helping you validate model capabilities in week one.
Original sourcesProduct Growth
02 / main
OpenAI engineers reveal agent loop optimization techniques
The takeaway: OpenAI optimizes ChatGPT and Codex agent performance by reducing redundant work across the harness, API, and inference layers. By maintaining persistent WebSockets, reusing prompt prefixes, and running safety checks in parallel, the architecture minimizes network latency and token costs.
Concrete details
- Persistent WebSockets replace HTTPS connections, eliminating TCP and TLS handshakes on repeated agent tool calls.
- GPT-5.6 Sol with max reasoning outscored Fable 5 on coding benchmarks while costing less than half as much.
Why it matters for this reader: Understanding harness and API optimizations helps you design agent architectures that minimize latency and control token expenses in production features.
Original sourcesByteByteGo Newsletter
03 / main
Anthropic's first technical PM outlines eval-driven development
The takeaway: Dianne Penn shares how Anthropic scales model capabilities using eval-driven development loops and token management strategies. Having guided products from Claude 2 through Fable and Claude Code, she highlights how PMs can navigate unpredictable model behavior.
Concrete details
- Dianne Penn joined Anthropic in 2023 as its first technical product manager when product had five engineers.
- Pioneered eval-driven development loops while launching capabilities like computer use, Model Context Protocol, and reasoning.
Why it matters for this reader: Learning how frontier lab product managers manage token budgets and eval loops helps you structure product workflows for unpredictable LLM capabilities.
Original sourcesLenny's Newsletter
04 / main
How delivery platforms integrated LLMs into search architectures
The takeaway: DoorDash, Instacart, and Uber Eats took wildly different architectural paths to upgrade search. DoorDash uses LLMs to enrich offline knowledge graphs, Instacart fine-tunes Llama-3-8B for tail queries, and Uber Eats uses two-tower Qwen embeddings across all verticals.
Concrete details
- DoorDash achieved a 30% lift in carousel trigger rates by using LLMs as retrieval guardrails.
- Instacart expanded query rewrite coverage from 50% to over 95% using fine-tuned Llama models.
Why it matters for this reader: These implementation patterns show you how to match LLM search features with existing infrastructure, balancing real-time latency against offline background processing.
Original sourcesByteByteGo Newsletter
05 / main
Claude Opus 5 benchmark reveals brilliant performance and quirks
The takeaway: Hands-on testing of Anthropic's new Opus 5 model shows top-tier intelligence paired with notable personality quirks. While the model excels on complex tasks, its high verbosity and overly cautious behavior can create friction during live software development tasks.
Concrete details
- Opus 5 was evaluated across six blind tasks against six competing models, including GPT-5.6 Sol.
- The model exhibited neurotic behavior in live tests, refusing to execute a software merge conflict.
Why it matters for this reader: Evaluating model personality alongside benchmark scores ensures you select the right LLM for interactive user workflows versus background automation.
Original sourcesLenny's Newsletter
06 / main
Sam Altman shares ChatGPT workflows for product managers
The takeaway: OpenAI CEO Sam Altman published practical guidelines detailing how product managers can incorporate ChatGPT Work into their daily routines. The advice focuses on integrating conversational AI models directly into standard product planning and team workflows.
Concrete details
- Sam Altman published guidance specifically targeted at product managers using ChatGPT Work.
- The recommendations cover incorporating conversational models into day-to-day PM task execution.
Why it matters for this reader: Adopting executive-tested ChatGPT workflows allows you to speed up routine product management tasks and focus more time on high-judgment evals.
Original sourcesProduct Growth
Action items
- Create a 100-prompt cold-start eval set for your agent feature before writing product requirements.
- Audit your agent architecture to verify WebSockets and prompt caching are eliminating redundant payload transfers.
- Evaluate whether head search queries can be cached offline while routing tail queries to fine-tuned models.
See every issue behind this brief
Optimizing ChatGPT's Agent Loop
- How ChatGPT Optimizes its Agent Loop: Harness, API, and InferenceByteByteGo Newsletter · Included
Building Frontier-Lab Quality AI Evals
- How to Build Frontier-Lab Quality Evals with Daniel McKinnon, ex-PM at Meta, GoogleProduct Growth · Included
Product Lessons from Anthropic's First Technical PM
- Anthropic’s first technical PM on token maxing, the jagged edge, and living in the future | Dianne PennLenny's Newsletter · Included
How DoorDash, Instacart, and Uber Eats Integrate LLMs into Search
- Why DoorDash, Instacart, and Uber Eats Integrated LLMs Into Search Three Different WaysByteByteGo Newsletter · Included
Claude Opus 5 Evaluation and Benchmarks
- 🎙️ How I AI: Claude Opus 5 Review + Browser use in Codex + How Cursor and a Raspberry Pi makes AI funLenny's Newsletter · Read, not included
- Claude Opus 5 review: this model is brilliant (but annoying)Lenny's Newsletter · Included
NVIDIA Open Models
- How NVIDIA Builds Open Models for the Age of AIByteByteGo Newsletter · Read, not included
ChatGPT Workflows for Product Managers
- Ultimate Guide: ChatGPT Work For PMsProduct Growth · Included
Hardware Hacking with Cursor
- From zero coding background to hardware hacker: How Cursor + a Raspberry Pi makes AI funLenny's Newsletter · Read, not included
Community Wisdom Pilot Purgatory
Vibe Coding PM Interviews
- Vibe Coding Interviews for PMs: Complete GuideProduct Growth · Read, not included
Lenny's Product Pass
- 11 products I love, free for a year—the biggest Product Pass expansion in 2 yearsLenny's Newsletter · Read, not included
Airtable Bundle Addition
- I'm adding Airtable to my BundleProduct Growth · Read, not included
Idempotency and Delivery Semantics
- A Detailed Guide to Idempotency, Delivery Semantics, and DeduplicationByteByteGo Newsletter · Read, not included
About Product Growth
Should you add Product Growth to Scottie?
Ship better AI products with tested product-management and engineering workflows.[1][2][3]
Who it’s for
A product manager building AI features who needs practical guidance on agents, evaluation, distribution, and day-to-day product work.[1][2]
What you’ll find in it
The issues linked below include “The PM's Guide to Agent Distribution: MCP Servers, CLIs, and AGENTS.md”, “How to Build Frontier-Lab Quality Evals with Daniel McKinnon, ex-PM at Meta, Google”, and “Ultimate Guide: ChatGPT Work For PMs”. Open them to judge the publication in its own words.
- Byline
- Aakash Gupta[1]
How to read Product Growth with Scottie
Add its public feed for publicly available issues. If you subscribe to paid issues, forward those emails to your private @scottie.fyi address. Scottie does not need access to your inbox.
Scottie read 4 Product Growth items and included 2 for the brief. The sources covered different things, so Scottie kept their stories separate instead of forcing a connection.
Read three issues from Product Growth
Read the publication in its own words. Scottie keeps these original links attached; it does not replace the writing.
Sources and official links
- Product Growth official publication and public archive
- The PM's Guide to Agent Distribution: MCP Servers, CLIs, and AGENTS.md
- How to Build Frontier-Lab Quality Evals with Daniel McKinnon, ex-PM at Meta, Google
- Ultimate Guide: ChatGPT Work For PMs
- Lenny's Newsletter official publication
- ByteByteGo Newsletter official publication